Filtering of false positive microRNA candidates by a clustering-based approach

BMC Bioinformatics

Table 4 Results of the performance analyses of ProMirII-g and miR-abela using human, mouse and rat genome data.

Software	Statistics	Human	Mouse (All)	Mouse (Distinct)	Rat
ProMirII-g	# of predictions	690	656	640	615
	# of TPs	215	199	183	127
	# of FPs	475	457	457	485
	# of real miRNAs missed	25	46	29	19
	SE	89.58%	81.22%	85.92%	86.99%
	PPV	31.16%	30.34%	28.59%	20.65%
miR-abela	# of predictions	1036	915	901	646
	# of TPs	149	140	126	86
	# of FPs	887	775	775	560
	# of real miRNAs missed	91	105	86	60
	SE	62.08%	57.14%	59.43%	58.90%
	PPV	14.38%	15.30%	13.98%	13.31%

It should be noted that in the mouse genome, 25 of the pre-miRNAs are duplicated. In other words, only 212 mouse pre-miRNAs are distinct in the genome. To avoid overestimation of the performance of the software, we identify the duplicated ones and conduct two measurements. #, number; FP, false positives; TP, true positives; SE, sensitivity; PPV, positive predictive value.

ISSN: 1471-2105